What will your AI feature
cost to run?
Estimate your token bill before launch and compare 17 models from five providers on the same workload.
AI API cost calculator.
Enter your workload once, then compare it across 17 models.
- Per request
- USD 0.0090
- Per 1,000 requests
- USD 9.00
- Per day
- USD 9.00
- Per year at this rate
- USD 3,240.00
- Input · cached · output share
- 44% · 0% · 56%
- Tokens per month
- 75,000,000
Taxes, tool calls, images, audio and cache-write charges are not included.
Check Claude API pricing ↗Monthly cost comparison.
| Model | Tier | Per 1,000 requests | Monthly | Relative cost |
|---|---|---|---|---|
| OpenAI | Fast & low-cost | USD 0.45 | USD 13.50 | |
| Anthropic | Fast & low-cost | USD 0.45 | USD 13.50 | |
| DeepSeek | Fast & low-cost | USD 1.20 | USD 36.00 | |
| Fast & low-cost | USD 1.25 | USD 37.50 | ||
| Mistral | Balanced | USD 1.75 | USD 52.50 | |
| Fast & low-cost | USD 1.85 | USD 55.50 | ||
| Balanced | USD 3.38 | USD 101.25 | ||
| OpenAI | Fast & low-cost | USD 3.75 | USD 112.50 | |
| OpenAI | Balanced | USD 4.50 | USD 135.00 | |
| DeepSeek | Balanced | USD 4.62 | USD 138.60 | |
| OpenAI | Balanced | USD 9.00 | USD 270.00 | |
| Anthropic | Balanced | USD 9.00 | USD 270.00 | |
| Frontier | USD 9.00 | USD 270.00 | ||
| Frontier | USD 10.00 | USD 300.00 | ||
| Anthropic | Frontier | USD 18.00 | USD 540.00 | |
| Anthropic | Frontier | USD 36.00 | USD 1,080.00 | |
| OpenAI | Frontier | USD 45.00 | USD 1,350.00 |
Estimates only, from public USD list prices checked November 12, 2026 on each provider’s pricing page. Prices are standard-tier text rates. Cheaper is not the same as suitable: test quality on your own task before choosing. Token counts differ between tokenizers, so the same text can cost a different number of tokens on each model. Reasoning or thinking tokens are billed as output on the models that produce them. Regional processing, priority tiers, fine-tuning, tools and media are priced separately. Nothing you enter leaves your browser.
Measure a real request first.
Send a few realistic requests in a development environment and read the usage numbers the API returns. Include the system prompt, conversation history, retrieved documents and tool definitions in the input count. Those parts often outweigh the user’s message.
Use the token estimator for a quick first pass. Then confirm with the provider’s token-counting endpoint for the exact model you plan to use.
Read: estimate LLM costs before you build →Cache, batch, then route.
Put stable content such as instructions and reference text at the start of the prompt so it can be cached. Move work that can wait into a batch job or off-peak window. Send simple tasks to a smaller model and keep larger models for the cases that need them.
Set a monthly spending limit in the provider console and an application-level request ceiling. An estimate is a plan, not a cap.
Read: when prompt caching cuts costs →API cost calculator FAQ.
How is an AI API request priced?
Most providers charge separately for input tokens you send and output tokens the model generates, each priced per million tokens. Output usually costs several times more than input. Cached input, batch processing and long prompts can change the rate.
How many tokens is my prompt?
A common rule of thumb is about four characters, or roughly three quarters of an English word, per token. Tokenizers differ between models and languages, so measure real prompts with the provider’s token-counting tool before setting a budget.
Does the calculator include reasoning tokens?
Enter them as part of the output tokens. Models that produce thinking or reasoning tokens bill them as output, so a short visible answer can still use many output tokens.
What does the cache share mean?
It is the percentage of each request’s input that the provider serves from its prompt cache, such as a long system prompt reused across requests. Cached input is billed at a lower rate. Some providers also charge to write to the cache or to store it, which this estimate does not include.
Are these prices current?
The presets were checked on November 12, 2026 against each provider’s public pricing page and are linked beside each model. Prices change, and your account’s tier, region or agreement may differ, so confirm before committing a budget.